time_test fails when running after any test which uses strptime().

The bug can be easily reproduced by running test_time twice:

$ TZ=Europe/Kiev ./python -m test.regrtest -ugui -v test_time test_time
...
======================================================================
FAIL: test_strptime (test.test_time.TimeTestCase)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/serhiy/py/cpython-2.7/Lib/test/test_time.py", line 120, in test_strptime
    (format, strf_output))
AssertionError: conversion specifier '%Z' failed with 'MSK' input.

----------------------------------------------------------------------

All works on 3.x and with TZ=UTC.

May be this is related to issue13309.

If call time.localtime() with time before 1990-03-25T03:00, EET is turned to MSK.  I guess this is a time when the Europe/Kiev timezone was introduced.

>>> import os, time
>>> os.environ['TZ'] = 'Europe/Kiev'
>>> time.tzset()
>>> time.localtime(638319599)
time.struct_time(tm_year=1990, tm_mon=3, tm_mday=25, tm_hour=1, tm_min=59, tm_sec=59, tm_wday=6, tm_yday=84, tm_isdst=0)
>>> time.strftime('%Z', time.gmtime())
'MSK'
>>> time.localtime(638319600)
time.struct_time(tm_year=1990, tm_mon=3, tm_mday=25, tm_hour=3, tm_min=0, tm_sec=0, tm_wday=6, tm_yday=84, tm_isdst=1)
>>> time.strftime('%Z', time.gmtime())
'EET'

C function localtime() implicitly calls tzset() and sets global C variables tzname, timezone, and daylight, but these changes are not exposed as variables in the time module. C and Python ideas about timezone becomes different. This looks as a bug in time.localtime() and other functions that implicitly change timezone variables.